On 11/05/15 10:50, Yuko Katabami wrote:
On 05/11/2015 04:56 PM, Doron Fediuck wrote:
>
> On 11/05/15 09:20, Yuko Katabami wrote:
>> Hi all,
>>
>> I have another question on a 3.6 string.
>>
>> *File:***AppErrors
>> *Resource ID:***VAR__DETAIL__NOT_MEMORY_PINNED_NUMA
>> *String:***$detailMessage cannot accommodate memory of VM's pinned
>> virtual NUMA nodes within host's physical NUMA nodes.
>> *Question:* I have a bit of difficulty clearly understanding this
>> string. Particularly the part "memory of VM's pinned virtual NUMA
nodes"
>> Could anyone paraphrase this? (Could it possibly be "memory of VM's
>> pinned to virtual NUMA noteds"?)
>>
>>
>> Thank you,
>>
>> Yuko
>>
>>
>>
> Hi Yuko,
> NUMA is splitting memory+cpu into smaller cells, where each cell has
> cpu+ram. This is possible in real machines, and also inside the guest.
>
> The best performance is achieved when you make sure each virtual cell
> (virtual node / vnode) is pinned to a physical cell (pnode).
>
> If the physical node is smaller than the vnode, or occupied with other
> VMs, you will get this error.
>
> Doron
Hi Doron,
Thank you very much for your explanation.
I think I understand the concept now but I would like to double check
with you -- can it be re-phrased as:
"cannot allocate memory to the VM's virtual NUMA nodes which is pinned
to host's physical NUMA nodes."
(I also want to check accommodate vs allocate)
Hi Yuko,
this is about node (cell) capacity, so we should try and keep it simple.
Kind regards,
Yuko